Prometheus
DE EN

01 — Why now

Open source carries the value creation. It is not being measured.

Over 90% of modern software value creation relies on open source. Nevertheless, maintainers and foundations make decisions about roadmap, sunset, and funding largely based on intuition.

The blind spot

What is measured today — and what is not.

The established metrics are not wrong. They just measure something different than the questions being asked.

Reach NPM downloads, PyPI stats

Is measured — counts package transfers, including CI runs and mirror traffic.

Attention GitHub stars, watchers

Is measured — a star is a bookmark, not evidence of productive use.

Usage duration How long does the library actually run?

Is not measured systematically by anyone.

Activity & Depth Which features are used at all?

Is not measured systematically by anyone.

Prometheus Reach and usage duration, verifiably anonymous

Fills the two dashed lines — as Open Data, on multiple independent mirrors.

The bars are qualitative: they show whether a metric is captured, not how high it is.

Consequence

Three tools, three different answers to the same question.

NPM Downloads

Count packages that were downloaded. They say nothing about whether the library was ever executed — including CI runs and mirror traffic.

GitHub Stars

Count attention, not adoption. A star is a bookmark, not evidence of productive use.

Prometheus

Measures reach and usage duration — verifiably anonymous, as Open Data, on multiple independent mirrors, verifiable via a public audit log.

Who suffers from it

The blind spot has a concrete cost.

Maintainers decide without data

Whether to sunset a module or introduce a breaking change can only be estimated without usage data.

Foundations fund by visibility

Funding flows to where attention is — not necessarily to where critical infrastructure lies.

Vendors don't know their dependencies in depth

Without reach insight, supply chain risk remains a list of package names rather than a reliable assessment.

Research builds custom crawlers every time

Every study starts with the same data procurement — not reproducible, not comparable, not shared.